GitHub Copilot’s move toward token-based billing is a preview of a broader shift. As assistants become long-running workers, the old SaaS seat model becomes less descriptive. The buyer is no longer paying only for access. The buyer is funding compute, retries, context windows, tool calls, and review time.
For builders, the pricing lesson is to connect usage to completed work or avoided cost. Invisible token meters create anxiety unless customers can see the business outcome they purchased. That is especially true in vertical markets where margins are tight and AI budgets compete with labor, compliance, and field operations.
The strongest products will instrument the workflow, not just the model. They will show what was completed, what required human review, what risk was reduced, and where spend stopped being productive.
Grey Haven read: As agents do longer-running work, pricing has to map to business value rather than abstract access. Tokens, tool calls, retries, and review cycles become part of the workflow cost a buyer must understand.
Reader guidance: Treat the item as a signal about operating readiness rather than headline momentum. The useful question is which control, workflow, or evidence requirement must be strengthened before the same pattern reaches a production system with customers, regulators, or internal teams depending on it.
Operator guidance: Instrument completed work, avoided rework, escalation rates, and unproductive spend. Watch pricing models that expose raw usage growth without showing which tasks were finished or which risks were reduced.
Source/provenance: Grey Haven Signal update linear-ASM-94-2, based on the existing backfill draft and source detail Hacker News discussion: https://news.ycombinator.com/item?id=47923357.
